Bots Become Us

Mark Zuckerberg’s new goal is creating artificial general intelligence

“We have built up the capacity to do this at a scale that may be larger than any other individual company.”

The Verge

There are a lot of interesting tidbits in Mark Zuckerberg's interview with Alex Heath. Like, for instance, the awfully specific number drop "340,000 Nvidia’s H100 GPUs" that makes this all sound like a corporate nuclear arms race. But his closing thoughts stuck out to me in particular:

He sees a future in which virtual worlds are generated by AI and filled with AI characters that accompany real people. He says a new platform is coming this year to let anyone create their own AI characters and distribute them across Meta’s social apps. Perhaps, he suggests, these AIs will even be able to post their own content to the feeds of Facebook, Instagram, and Threads.

We've been building to this world for a while with digital characters and IP,¹ but the world envisioned by Zuckerberg above would expand that exponentially. Almost nine years ago I wrote (half jokingly) about the notion of "bots thanking bots", also in the context of Facebook:

Which leads to the next question: at what point do bots start talking to bots? You know, why should you have to type “thank you!” when you can reply to a text with “1”? Or better yet, why should you have to type the “1” at all? If Facebook knows you want to say “thank you” to everyone (bots included) who wished you a happy birthday, shouldn’t they just give you the option to let Facebook do that for you on your behalf?

And that leads to the notion of having Facebook automatically say “happy birthday” to a friend on their birthday each year. If you can do that and then the Facebook “thank you” bot can reply to the “happy birthday” bot, we would have some hot bot-on-bot action.

We’re just now getting used to the first layer of interacting with bots for various services. But having bots chat with other bots is the next logical step that probably isn’t that far off. In many ways, it may be easier to make happen because it removes the flawed human variable in the equation. I’m both kidding and entirely not kidding.

This is definitely already happening on Xitter, but it's largely stupid spam bots just trying to game the algorithms for whatever purpose (and Xitter saying they're going to combat this and seemingly doing nothing against it as it seems far worse now than ever before). When Meta (and others) start to productize the creation of these bots at scale, things are going to get even more weird.

It all sounds fairly dystopian, but there's undoubtedly some entertainment value here. The question is if there's any actual value after the novelty wears off. Most of these AI chat bots seem to be fun for about five seconds, notably Meta's faux celebrity bots. Part of the issue there is obvious: you know you're not talking to the actual celebrity and perhaps all of this is only interesting if you at least think you are talking to an actual person.

Why? Because it's both human nature and the stakes are higher. There are real world ramifications based on what you say to an actual person – which is part of what makes social media discourse, as it has spiraled downward, crazy: you may be typing on a machine, but you're talking to a real person, with real emotions, who will take those back in to their own real world! – with bots, not yet at least.

So step one is probably making you truly believe you're talking to an actual person. There are moral and very gray lines there, and we're going to cross them very soon if we haven't already. Step two is the notion of 'bots as bots' becoming more compelling, per above. Probably as they become more human-like. Yes, like Samantha in Her. This is at least part of what made Microsoft's "Sydney" hallucination so fascinating a year ago. Step three sounds too wild right now but you have to believe where this is heading unless we go out of our way to stop it for various reasons: those bots taking on a human-like form to be able to fully "exist" in the aforementioned real world with aforementioned real world stakes. Yes, like Dolores in Westworld.²

¹ The company I linked to, Superplastic, is one I invested in via my role at GV, partially on the thesis of digital IP starting to move in these directions...

² Or Ava in Ex Machina...